There is no single "best" way to compress a PDF. A text-only report, a scanned contract and a 40-slide presentation full of photos get heavy for different reasons, so each one shrinks best with a different method. Pick the wrong method and you either save almost nothing or end up with a file that is small but blurry, unsearchable, or missing its links.
This article compares the methods side by side and tells you which one to use for which file. If you want the theory first (lossless versus lossy, why PDFs get big), read What Is PDF Compression?. If you already know what you want and just need the clicks, see How to Compress a PDF.
Quick answer
For a scan or a file made of photos that has to fit an upload limit, rasterizing the pages (the "Strong" mode in Compress PDF, quality around 60%) gives the biggest saving, at the cost of selectable text. For a text document you need to keep searchable, use the "Lossless" mode, but expect a small gain. For a heavy Word or PowerPoint export, the most effective fix is usually to export it again from the source with smaller image settings. Deleting pages you don't need, or splitting the file, works for any type.
First, find out what makes your file heavy
Choosing a method is much easier once you know where the bytes are. You don't need special software for a rough diagnosis; three quick checks are enough.
- Size per page. Divide the file size by the number of pages. A text page exported from Word typically weighs tens of kilobytes. If you see several hundred kilobytes or more per page, there are images inside, either photos or whole scanned pages. PDF Info shows the page count and file size together, along with the "Producer" field, which often names the scanner app or the program that created the file.
- Can you select the text? Open the PDF and try to highlight a sentence. If you can, the text is real text. If the whole page gets selected like a picture, or nothing happens, it is a scan or a photo, and the text is just pixels.
- Where did it come from? A scanner, a phone camera, an export from Word or PowerPoint, or a design program. The origin usually tells you whether you can go back to a source file, which is often the best option of all.
In practice, most oversized PDFs fall into one of three groups: scans and phone photos (each page is a large image), documents with embedded photos (text plus heavy pictures), and files that are already lean text where little can be gained.
The two methods SmokePDF's compressor uses
It's worth being precise here, because many online tools describe themselves vaguely as "smart compression". The Compress PDF tool on this site does exactly two things, both inside your browser, without uploading the file.
1. Structure rewrite ("Lossless — keeps text selectable")
The tool reads the PDF and saves it again with a more compact internal layout: the many small objects that describe the document are packed into compressed "object streams". Text, fonts and images are left exactly as they were, so nothing you can see changes and the text stays selectable and searchable.
The trade-off is the size of the gain. On a file that came out of a modern program, the saving is often just a few percent, and sometimes there is nothing to gain at all. In that case the tool tells you the file is already compact and gives you back the original rather than a "compressed" copy that isn't smaller. Older files, or files assembled by basic software, sometimes shrink more noticeably.
2. Whole-page rasterization ("Strong")
This mode draws every page as a picture and stores it as a JPEG at the quality you choose on the slider (10–90%, default 60%). At higher quality settings the page is also drawn at a higher resolution: roughly 144 dpi at 70% and above, about 108 dpi between 40% and 69%, and about 86 dpi below that. That is fine for reading on a screen and for most uploads, but not for professional printing.
The saving on scans and photo-based files is usually large. In one of our tests, a 6-page scanned file of 7.33 MB came out at about 800 KB at 60%, still readable. But the cost is real: the text becomes part of the image, so you can no longer select, copy or search it, and links, form fields and bookmarks do not survive, because each page is rebuilt from scratch as a picture. If the rasterized version happens to be larger than the lossless one (which can happen with lean text files), the tool keeps the lossless result instead.
What the tool does not do: it does not pick out individual images and shrink only those, it does not remove fonts or metadata selectively, and it does not delete attachments embedded inside the PDF. Those methods exist, but they need other software.
Methods that need other software or another tool
3. Selective image recompression and downsampling
This is the "surgical" method: the software finds each embedded image, lowers its resolution (for example to about 150 dpi for screen or 300 dpi for print) and re-encodes it, while leaving the text and vector graphics untouched. You get much of the saving of rasterization while the text stays real text. It is the ideal method for a report or brochure full of photos.
You'll find it in the optimization or "reduce file size" features of desktop PDF editors, and in the open-source command-line tool Ghostscript, whose presets (such as /screen, /ebook and /printer) trade resolution for size. Results depend heavily on the settings, so check the output before sending it.
4. Removing unneeded content
Some PDFs carry weight you never see: fully embedded fonts when a subset of the characters used would do, attached files, page thumbnails, editing data saved by the creating program, or long metadata. Desktop optimizers can strip these one by one. The gain varies a lot: large on a file with an embedded video or full font families, close to zero on a clean export.
The simplest form of this method, removing whole pages, is something you can do here. Remove Pages copies the pages you keep into a new PDF without touching their content, so there is no quality loss. Note that bookmarks, form-field structure and document metadata are not carried over to the new file. Dropping a blank scan page, a duplicate or an appendix nobody asked for is often the cheapest saving available.
5. Grayscale conversion
Removing color can help when the colors themselves are the weight, as in color scans of black-and-white documents. Desktop optimizers can convert embedded images to grayscale without touching the text.
Be aware of how the Grayscale PDF tool on this site works: it redraws each page as a high-quality image so it can remove the color. That means the text becomes an image and the file is not necessarily smaller; on a text document it may even grow. Use it when a portal or a printer asks for black and white, not as a size tool on its own. If size is the goal, Strong compression is the more direct route.
6. Splitting the file
When a portal limits the size of each upload rather than the total, splitting may be the only option that loses nothing. Split PDF cuts the file every N pages and gives you the parts in a ZIP. Each part keeps its original quality. This doesn't help when the portal wants a single file.
7. Re-exporting from the source
If the PDF came from Word, PowerPoint or a similar program and you still have the original, this often beats any compression done afterwards. Before exporting, reduce the images in the source (in Microsoft Office, "Compress Pictures" lets you choose a lower resolution for all images), then choose the smaller output option when saving as PDF. Word on Windows, for example, offers "Minimum size (publishing online)". If you embed fonts, the option to embed only the characters used keeps them light. LibreOffice's PDF export has similar settings for image quality and resolution. The result keeps real, searchable text with lighter images.
Comparison table
The savings below are described qualitatively on purpose: the real figure depends entirely on what is inside your file.
| Method | Typical saving | What you lose | Best for |
|---|---|---|---|
| Structure rewrite (SmokePDF Lossless) | Small, often a few percent; sometimes none | Nothing visible | Text documents, signed or official files, a first safe try |
| Whole-page rasterization (SmokePDF Strong) | Large on scans and photos; can be small or negative on lean text | Selectable text, search, links, forms, bookmarks; some sharpness | Scans and phone photos going to an upload limit |
| Selective image downsampling (desktop software) | Medium to large on image-heavy files | Image detail, depending on settings; text stays real | Reports, brochures and portfolios with many photos |
| Removing unneeded content (desktop optimizers; Remove Pages here) | Varies from none to large | Whatever you remove; with Remove Pages, also bookmarks and form structure | Files with extra pages, attachments or full embedded fonts |
| Grayscale | Modest at best; with SmokePDF's tool, not reliable as a size method | Color (stamps, signatures in blue ink, charts) | Color scans of black-and-white documents |
| Splitting | None in total; each part is smaller | The convenience of a single file | Portals with a per-file limit |
| Re-export from source | Often large on Office files with photos | Image resolution you chose to lower | Word and PowerPoint documents you still have the original of |
Which method for which situation
An email attachment
Many email services cap attachments somewhere around 20–25 MB, and recipients on mobile appreciate much less. If the file is a scan, Strong compression at 60% is usually enough. If it's a text document with photos and the recipient needs to copy text from it, re-export from the source or downsample images with desktop software instead of rasterizing. If it's lean text and still too big, it's probably very long: send only the relevant pages.
A government or university portal with a strict limit
This is where people most often face limits like 1, 2 or 5 MB, usually for scanned IDs, certificates and signed forms. These files are images anyway, so rasterization costs nothing important: use Strong, start at 60%, and go down to 40% only if needed, then zoom in on numbers and stamps before uploading. Remove blank or duplicate pages first. If the portal accepts several files, splitting is an alternative that keeps full quality.
Archiving
For files you'll keep for years, quality and searchability matter more than a few megabytes. Use lossless methods only: the structure rewrite, removal of truly unneeded pages, and re-exporting from the source if it's available. Avoid rasterization, which destroys the text layer for good. If an archived scan needs to be searchable, that requires OCR, which is a separate process explained in What Is OCR?.
Printing
Don't compress a file headed to a print shop unless the printer asks you to. Strong mode's resolution (at most about 144 dpi) is well below the roughly 300 dpi commonly used for print. If the file must be lighter, re-export from the source with print-quality settings, or downsample images no lower than about 300 dpi with desktop software.
A presentation or report full of photos
Selective downsampling or re-exporting with "Compress Pictures" is the right tool here: the photos get lighter and the text stays text. Rasterizing a 40-slide deck works, but the result can't be searched or copied from, which the reader will notice.
Combining methods in the right order
Methods can be combined, but the order matters:
- Remove first. Delete the pages you don't need before anything else. There is no point compressing pages you're about to throw away.
- Go back to the source if you can. A good export beats post-processing, and it keeps the text intact.
- Try lossless before lossy. If the structure rewrite already gets you under the limit, stop there.
- Compress strongly only once, from the original. Rasterizing an already rasterized file adds JPEG artifacts each time. If 60% isn't enough, go back to the original and try 40%, rather than compressing the compressed copy again.
- Split last, only if the target accepts several files and nothing else gets you under the limit.
When to think twice before compressing
- Digitally signed PDFs: any rewrite, even lossless, changes the file's bytes and can invalidate a digital signature. Send the signed original if its size is acceptable.
- Fillable forms: Strong mode turns the form into flat images; the fields are gone. Fill and submit the form first, or use lossless only.
- Documents someone needs to copy from: rasterization removes selectable text. A colleague who has to quote from your report will have to retype it.
- Files that are already small: if a 300 KB file won't go lower, there is often nothing worth removing. Compressing harder only damages it.
Frequently asked questions
Why did Lossless mode barely change my file?
Because most modern programs already save PDFs efficiently, so rewriting the structure finds little to pack. If the file is heavy because of images or scanned pages, only a method that touches the images (Strong mode, downsampling or re-export) will make a real difference.
Can I get the selectable text back after Strong compression?
Not from the compressed file itself, since the text is now part of a picture. Keep the original: it still has the text. If you only have the compressed version, OCR software can recognize the text again, with some errors depending on quality.
Is it safe to compress a PDF that contains personal documents?
With this site's tools, the processing runs in your browser and the file isn't uploaded to a server, which is how we can handle ID scans without storing them. With any other online service, check its privacy policy before uploading sensitive files.
Which quality should I pick on the Strong slider?
Start at 60%, which is the default and suits most scans. If the file is still too big, try 40%; if small print or stamps become hard to read, try 70%. Always re-run from the original file rather than compressing the compressed one.
Does compression remove hidden data like the author's name?
Not reliably. The lossless mode rewrites the file but doesn't selectively strip metadata. Rasterizing builds new pages, but you should still not count on any compressor as a privacy tool; a dedicated metadata-removal feature in desktop software is the safer route.
The bottom line
Match the method to what makes the file heavy. Scans and photos: rasterize once, at moderate quality. Text with photos: shrink the images, ideally from the source, and keep the text real. Lean text: rewrite the structure, remove pages, or split, and accept that the gain will be small. When the document matters long term, prefer lossless every time.
